Teleworking has been linked to perceived autonomy, job satisfaction, organizational commitment, wellbeing, and to lower stress and burnout (e.g., Baruch, 2000; Charalampous et al., 2019; Felstead and Henseke, 2017; Fonner and Roloff, 2010; Gajendran and Harrison, 2007; Redman et al., 2009; Vega et al., 2015). Flexibility in where and when to work may increase perceived autonomy, which is associated with job satisfaction and performance (Beauregard et al., 2019; Gajendran and Harrison, 2007; Gajendran et al., 2015). Higher productivity or performance has been found for both self-rated (Delanoeije and Verbruggen, 2020; Vega et al., 2015), supervisor-rated (Gajendran et al., 2015; Kossek et al., 2006) and objective measures (Bloom et al., 2015; Gajendran and Harrison, 2007). Proposed mechanisms include longer working hours (Baruch, 2000; Felstead and Henseke, 2017), for instance by using time saved from commuting (Beauregard et al., 2019), or greater efficiency at home potentially due to less distractions (Baruch, 2000; Bloom et al., 2015) or increased autonomy (Gajendran et al., 2015).

Evidence is not uniform. Some studies found no relationship between productivity and teleworking (Aguilera et al., 2016), and concerns have been raised about reduced knowledge transfer and weaker co-worker relationships (Beauregard et al., 2019). Such negative effects might depend on intensity, as only high-intensity telework was associated with worse co-worker relationship quality in Gajendran and Harrison (2007). Likewise, the effects of teleworking on job satisfaction and professional isolation have been found to vary with telework extent (Golden and Veiga, 2005; Golden et al., 2008).

Teleworkers may also fear being “out of sight” for task allocation and promotion (Beauregard et al., 2019). Although this was not supported in the Gajendran and Harrison (2007) meta-analysis, experimental evidence suggests the concern may be warranted: call-center employees randomly assigned to telecommute were more productive, more satisfied, and less likely to leave, yet slightly less likely to be promoted when controlling for productivity (Bloom et al., 2015).

Teleworking has also been associated with experiencing more positive affect and less negative affect on teleworking days as compared to office days (Anderson et al., 2015). This association varied depending on individuals' openness, rumination, and social connectedness outside the workplace (Anderson et al., 2015). Part-time teleworkers reported higher concentration and lower need for recovery on home vs. office days (Biron and Veldhoven, 2016). Redman et al. (2009) found that homeworking positively impacted employee wellbeing, through higher positive affect, empowerment, job and life satisfaction, and lower stress and burnout.

Findings on work-life balance and well-being are mixed. Some studies find improved balance and lower overall work-family conflict (e.g., Fonner and Roloff, 2010; Gajendran and Harrison, 2007). However, work-life balance can include both how family interferes with work (family-work conflict), and how work interferes with family (work-family conflict). Redman et al. (2009) found that more hours worked from home reduced family-work conflicts, but was not associated with work-family conflicts. Golden et al. (2006), however, found that more teleworking was associated with higher family-to-work conflict and lower work-to-family conflict. Others have found adverse effects such as overworking, increased spillover between home and work, and increased stress related to balancing work and domestic chores (Eddleston and Mulki, 2015; Felstead and Henseke, 2017; Kazekami, 2020). Grant et al. (2013) found that while many reported improved work-life balance, many also mentioned overworking and an inability to completely switch off from work and lacking recuperation, with adverse impact on wellbeing. Differences across workplaces, work cultures (Beauregard et al., 2019), and telework extent may contribute to inconsistent results (Allen et al., 2015).

Variation in findings may also reflect differences in study designs and recruitment strategies. Working from home has been examined through surveys of employees (e.g., Abujarour et al., 2020; Baruch, 2000; Eddleston and Mulki, 2015; Gajendran et al., 2015; Redman et al., 2009) or managers (e.g., Gajendran et al., 2015; Redman et al., 2009) or qualitative interviews (e.g., Eddleston and Mulki, 2015; Grant et al., 2013; Purwanto et al., 2020). Studies range from single workplaces (Anderson et al., 2015; Bloom et al., 2015; Delanoeije and Verbruggen, 2020; Eddleston and Mulki, 2015) to multiple businesses (Baruch, 2000; Redman et al., 2009), national panels (Kazekami, 2020), national survey series (Felstead and Henseke, 2017) or snowball sampling (Biron and Veldhoven, 2016; Fonner and Roloff, 2010).

Most studies are cross-sectional, although some used quasi-experimental designs (Bloom et al., 2015; Delanoeije and Verbruggen, 2020). These designs also have limitations, including employer selection into telework eligibility (Delanoeije and Verbruggen, 2020), and random assignment among volunteers only (Bloom et al., 2015), potentially limiting generalisability beyond those who opted in.

During the Covid-19 pandemic, employees were strongly encouraged or required to work from home, which may reduce some positive associations observed in studies of voluntary teleworking. Pandemic teleworking also introduced constraints such as inadequate home workspaces, insufficient internet or equipment, and increased demands from remote schooling and childcare. Some studies report negative associations between pandemic teleworking and wellbeing (Kornadt et al., 2025; Nordenmark and Vinberg, 2024).

Other studies report increased job satisfaction, efficiency and concentration, improved work-life balance, and lower burnout and work-related stress (Baert et al., 2020; Tao et al., 2023; Wong and Cheung, 2020). Overall, impacts on work-life balance and mental health appear heterogeneous and related to factors such as gender, age, occupation, children in the home, home office setup, and personality traits and preferences (Allen et al., 2021; Elbaz et al., 2023; Hall et al., 2023).

A similar pattern of heterogeneity is observed for the impact on work performance (Elbaz et al., 2023; Hall et al., 2023). Some studies find similar or higher self-reported productivity/performance when teleworking compared to office work (Beck et al., 2020; Deole et al., 2021; Etheridge et al., 2020; Innstrand et al., 2022), but results differ across industries, occupations, and socioeconomic groups (Etheridge et al., 2020). Reduced performance has been found for low earners and self-employed people (Etheridge et al., 2020), as well as for women with children (Deole et al., 2021; Etheridge et al., 2020). In contrast, having an appropriate workspace has been associated with better productivity (Awada et al., 2021; Beck et al., 2020), and supervisor or organizational support is linked to greater engagement and supervisor-rated performance (Katsaros, 2025).

Qualitative evidence highlights additional trade-offs. Interviews with Indonesian teachers revealed benefits such as flexibility and reduced commuting stress and costs, but also increased electricity/internet costs and potential data security issues (Purwanto et al., 2020). Family-work conflict can negatively influence attitudes toward working from home (Abujarour et al., 2020). Positive attitudes, in turn, are associated with higher productivity. Perceived technology usefulness predicts attitudes, and this relationship is moderated by sufficient internet speed (Abujarour et al., 2020).

Some studies have found that people enjoy working from home and would like to telework more in the future (Beck et al., 2020), with most desiring 1 or 2 days per week (Wong and Cheung, 2020). Some of the negative associations found between working from home and work-life balance and mental health may in part be due to other pandemic restrictions that co-occurred, such as limits on social and organized activities (Elbaz et al., 2023). Indeed, longitudinal studies found that while working from home was positively associated with loneliness (Fostervold et al., 2024) and negatively associated with wellbeing (Kornadt et al., 2025) in the initial stages of the pandemic, these effects were weaker or not related as the time went on.

While general wellbeing captures more stable evaluations of life and work, our interest lies in how day-to-day variation in work location relates to immediate experiences. The telework literature has predominantly examined broader indicators of wellbeing (e.g., job satisfaction, mental health, work-life balance), with comparatively less attention given to transient, day-specific affective states (Blank et al., 2023; Castro-Trancón et al., 2024; van Dick et al., 2024). Given our focus on within-person day-to-day variation, we therefore examine same-day mood as a more transient construct that can vary in response to daily scheduling choices between home and office work.

Research on telework has operationalized mood and work performance using different instruments and time frames. Some studies draw on established multi-item scales (Anderson et al., 2015; Delanoeije and Verbruggen, 2020; Etheridge et al., 2020), whereas others use single-item indicators (Kazekami, 2020). Temporal framing also varies, from retrospective evaluations (e.g., Felstead and Henseke, 2017), to assessments of the current day (e.g., Delanoeije and Verbruggen, 2020) or general life satisfaction (e.g., Redman et al., 2009). Such differences are consequential when the objective is to examine day-level experiences.

Mood is a transient state, and retrospective reports are therefore vulnerable to recall biases, including mood-congruent memory and duration effects (Ebner-Priemer and Trull, 2009). Same-day assessments are better suited to capturing everyday affective fluctuations. Repeated measurements across waves further allow for the examination of within-person variation over time (Curtis et al., 2020), but require limiting respondent burden. Brief, conceptually distinct items are therefore preferable in longitudinal designs (Fisher and To, 2012; Wilhelm and Schoebi, 2007).

Work performance has likewise been measured with instruments differing in recall period and scope (Koopmans et al., 2013) (Ma and Ye, 2019). When analyzing day-level associations between telework, mood, and perceived performance, measures referring to the same temporal window are methodologically more appropriate than broad retrospective evaluations.

Accordingly, the present study assesses mood as a same-day state using a brief set of bipolar items, and subjective work performance with reference to the current day. Both constructs were measured repeatedly across survey waves. The specific operationalizations are described in Section 2.3.

Teleworking has been associated with both advantages (e.g., flexibility, reduced commuting time, improved work-life balance) and challenges (e.g., social isolation, blurred boundaries between work and personal life, unequal access to suitable home-working conditions). However, the evidence stems from two structurally different teleworking regimes. Before the Covid-19 pandemic, teleworking was largely voluntary and implemented within organizations already supportive of remote work. Much of the evidence therefore reflects the experiences of employees who opted into teleworking (Allen et al., 2015), raising concerns about selection bias and limiting generalizability.

During Covid-19, teleworking was frequently externally driven and implemented under additional constraints, including home-schooling, limited workspace, and broader social restrictions. The literature reports mixed effects on wellbeing, work-life balance, and performance (Elbaz et al., 2023; Hall et al., 2023). Overall, pandemic findings resemble pre-pandemic results in some aspects but show a tendency toward more negative impacts on wellbeing and mental health. Some benefits and drawbacks are pandemic-specific (e.g., infection risk reduction and children at home), while others (e.g., reduced commuting time and increased autonomy) are more generalizable.

Despite these contextual differences, much of the research in both periods relies on cross-sectional designs (Elbaz et al., 2023), with few longitudinal studies that cover both teleworking and office work (Kornadt et al., 2025). Hence, there remains a need for longitudinal within-person studies that compare the same individuals when working from home and when working at the office. In addition, many studies rely on retrospective (e.g., Felstead and Henseke, 2017) or global measures of wellbeing and productivity (e.g., Redman et al., 2009). Hence, there is a need for instruments and research designs that capture the transient nature of daily mood and work performance.

This study addresses these gaps by using a multi-wave panel survey to repeatedly assess individuals' experiences of teleworking during the Covid-19 pandemic. By collecting within-person measures during the workday, we capture short-term fluctuations in mood and subjective performance associated with working from home vs. traveling to work.

As teleworking alters the daily structure of work, by removing commuting, reshaping spatial boundaries and modifying routines, its consequences are likely to emerge first in short-term affective states and task-related experiences rather than in broader evaluations of wellbeing. Examining these proximal responses provides a basis for understanding how repeated day-level differences may, over time, shape longer-term wellbeing and productivity. In increasingly hybrid work arrangements, such short-term dynamics are particularly relevant for organizational design.

The specific research questions are:

As governmental restrictions were gradually lifted, variation in teleworking differed across time and regions, creating a hybrid context in which employees switched between home and office work. We exploit this exogenous time and regional variation to reduce one source of selection bias: the choice to work from home.

Due to the rapid increase of Covid-19 cases, the Norwegian government declared a national crisis on the 12th of March 2020. All schools and kindergartens, libraries, museums and sport facilities were shut down, and several restrictions were imposed on traveling and social gatherings. WFH was imposed or strongly encouraged for those who did not hold a “necessary or critical job.” More restrictions on pub and restaurants service were imposed in the capital, Oslo. The government also released several crisis economic packages to support businesses through loans and avoid layoffs and bankruptcies. Nevertheless, at the beginning of April 2020, the unemployment rate was skyrocketed leaving more than 400 000 people (about 15% of the labor force) completely or partially unemployed (Høgseth et al., 2020).

By the end of April things started to re-open in Norway, starting with kindergarten and elementary schools as well as with dentists, psychologists, physiotherapists and similar. Through May and June most activities started again, people started to commute again, and also international travel restrictions were gradually and partially lifted. It seemed that “new normal” was established. There were still some restrictions for social gatherings, and most workplaces established a mix between commuting and teleworking. This situation lasted until September, when a second wave of Covid-19 infection struck, and many restrictions were quickly reintroduced.

The data used in the analysis were collected through an online survey in the spring, summer and fall of 2020 in Norway. We used a social media post for recruitment to reach a broad and diverse sample efficiently during the pandemic. Participants were recruited through a Facebook post which was shared on the Transport Economic Institute (TØI) website the 6th of May 2020 and advertised for 5 days, targeting adults in Norway. The survey was also promoted in a news report on the national broadcast (NRK) on the 18th of May.

A total of 2,348 people answered the first questionnaire (the baseline, also referred to as T0) between the 6 and the 25th of May. We cleaned the data by removing people who did not want to or were not able to participate in the follow-up surveys and ended up with 2,031 people. They were selected and contacted six more times in May, June, August and September 2020. Three follow-up surveys were sent between May and June, and another three during August–September. Table 1 shows the response period for each wave and the number of respondents.

Participation to the surveys was incentivized with lotteries: one respondent could win a universal gift card of NOK 1,500 for responding to each wave and their chances improved with the number of follow-up surveys completed.

Before the first three follow-ups in the spring, and in the fall, a preparatory e-mail was sent to all respondents with some instructions. In order more homogeneous answers and to strengthen the connection between mood and performance measurements with their working situation (i.e., commute or working from home), we instructed participants to answer at least 1 h after starting work, but before lunch. The instructions also informed them that they would receive the invitation to the survey at 9AM the following morning. Such communication was meant to increase compliance and remind them to check their spam folder if they did not receive the invitation. This instruction was repeated in all survey invitations and reminders. Most respondents complied with instructions: From the first follow-up, more than 90 percent report taking the survey when they had “worked a bit” or they were “about half-way” (while in T0 this share reaches only 46 percent).

Participation was entirely voluntary, and no personal data were collected through the social media platforms used for recruitment. The post and related news media articles directed potential participants to an external survey platform, where personal data were collected. These included online identifiers and self-reported health-related information and were processed solely for research purposes. Full written informed consent was obtained electronically prior to participation, including explicit consent for the processing of special categories of personal data. The project was approved by the relevant ethics and data protection authority (SIKT, ref. 874531). SIKT is the national service provider for the Norwegian knowledge sector, responsible for delivering infrastructure, services, and governance support related to research ethics, data protection, and secure data processing. The Institute of Transport Economics (TØI) acted as data controller for the study.

Before presenting the main results, we describe our sample in more details. After cleaning, we obtained a sample of 2,031 people who answered the baseline survey. About 54% (1,091 people) answered the survey all 7 times, while 14% answered 3 times or less. On average respondents answered 5.8 times out of 7.

Since many people could not work or were laid off during lockdowns, we limited our sample to people who were currently working in order to answer our research questions. Table 2 shows the summary statistics for the main characteristics of the sample.

About 70% of our responders are women, living in the east part of the country and are on average 43 years old (min 19, max 73). About 65% of respondents have kids. Of those, 27% have one child, 38% have two and 11% have more than two children living in the household. The remaining 24% have children older than 18 years old or not living in the same household.

Because of the recruitment condition, all respondents are either working or students (2%). About 46% report working in an office, 32% in academia or 11% being leaders. About 2% work in sales or service, 2% works in health services and less than 1% work in other sectors, such as farming, fishery, media or transport. About 36% report earning more than 67 thousand EUR¹ (where the average annual income in Norway is around 54 thousand EUR²) and around 60% have 5 or more years of university education.

The majority (51%) of respondents work within 10 km of their home and about 52% report a maximum 30 min journey to reach their workplace. About 48% of people express having flexible travel times. As a part of their commute, 25% of respondent report bringing children to school/kindergarten, 19% chain their trip with grocery shopping and 6% with performing activities such as exercise.

In order to reduce carry out analysis on comparable data, we restrict our sample to those who have a type of work that can be done from home. In particular, we exclude the 2% who report working in health services and the 1% who work in other sectors, such as farming, fishery, media or transport.

Since our main outcome variables (mood and performance) can be affected by various factors (such as tiredness and post-lunch dip) during the day, we further restrict our sample to those who complied with instructions and answered the survey when they had “worked a bit” or they were “about halfway” (83% of the sample).

After describing the data, we employ two methodologies each serving a different purpose. First, we address our research questions 1. and 2. by estimating the effect of working from home (versus traveling to work) on daily mood and work performance exploring the repeated nature of our survey. We use Generalized Least Squares (GLS) regressions for panel data³ and focus on individual fixed effects (including a dummy variable for each individual) model to control for all time invariant individual characteristics.

This study exploits the unique context of the Covid-19 pandemic, during which large shares of workers were required or strongly encouraged to work from home. Our identification strategy relies on the fact that the prevalence of working from home varied both over time (Figure 1) and across places of residence (Figure 2), as regional restrictions differed in timing and intensity. We use this exogenous variation to estimate the impact of working from home on daily mood and subjective work performance, reducing biases that typically arise when telework is a voluntary choice. Exploiting the fact that each respondent has answered our surveys several times and that the majority of the sample report both traveling to work and working from home, we can thus use this variation to estimate within person association of telework with the dependent variables.

Importantly, respondents also reported whether working from home on the survey day was freely chosen or imposed/encouraged by their employer. We thus conduct an additional robustness analysis excluding individuals who reported freely choosing to work from home (Step 2 in Table 3), thereby focusing on workers whose telework arrangement was externally driven.

Finally, we carry out GLS panel regressions with individual random effects on the sample of people reporting working from home that day. This analysis helps us answer research question 3, and further explore whether some individual characteristics moderate mood and performance outcomes specifically under teleworking conditions. Because individual traits are time-invariant, RQ3 is examined using between-person variation among home-working days. Our results report stepwise regression tables to show how our estimates are reliant on the choice of regressors and to illustrate the robustness of our findings.

In the baseline survey, we mapped several respondents' demographics and characteristics. All respondents are either working or studying, as this was a selection criterion (2% are students). 40% report working in an office, 35% in academia, 11% are leaders and only 2% respectively works in sales or service and as health workers. We observe that 74% of our sample report having at least some experience with working from home, while 532 people (26%) say they never worked from home before Covid-19 related restrictions (March 2020). From March to May 2020, about 88% of respondents have (exclusively) worked from home, while only 3% of our sample report never had worked from home in this period. About 80% of our sample also say WFH works quite or very well for them. Almost 90% report working similarly to how they worked before Covid-19, while 4% say they work more and 3% they work less than before.

Respondents also reported whether they worked from home or traveled to work “today,” each time they took the survey. About 9% (183) of respondents report always traveling to work, while 35% (712 people) reports always working from home the days they answered the survey. The rest of the sample, about 56%, reports both working from home and traveling to work in the days they answered the survey between May and September 2020.

Figure 1 shows that the number of those who work from home decreases over time (from over 84% to 58%) and increases slightly again in September. This is likely a consequence of the lifting of Covid-19 related restrictions through spring and summer 2020. In Norway, Covid-19 restrictions were introduced again after the summer, gradually and with some important difference between regions, see Figure 2. Initially the amount of people who freely chose to work from home was low (8%, of those working from home), while the majority (56%) had it imposed upon or (36%) strongly encouraged (Figure 1). Over time we see this imposition decreased and the amount of people choosing to work from home increased. People report their experience of working from home today to be quite good or very good 85% of the times.

In this section, we focus on the two main dependent variables: mood and subjective work performance. These are index variables spanning from 1 to 7 as described in Section 2.3. We address our main research question by studying the effect of working from home (vs. traveling to work) on self-reported immediate mood and subjective work performance reported at the beginning of the workday.

To graphically show our result in a simple way, we exploit the longitudinal nature of our data and select only people who report having changed between working from home and travel to work during the period of observation (N = 1136). We report mood and subjective performance averages within individuals who over time change from working from home to travel to work and vice versa. Figure 3 shows that the mean score when working from home is higher than when traveling to work for both indexes.

The average mood score is 4.5 for travelers and 4.6 for working from home. Average performance score changes from 4.6 to 4.7 when changing between traveling and working from home. Paired-samples t-tests were conducted to assess whether the changes were significant. We find a significant difference for mood (t(1136) = −3.8, p = 0.0001) and performance (t(1136) = −2.42, p = 0.016).

As described in Section 2.6 we run individual fixed-effect regression to study the impact of working from home on mood and performance within an individual exploiting its variation over time and place. In this way we are estimating an effect net of any individual characteristic, which may affect whether or not respondents work from home, or how they score in mood and performance, removing the effect of time-invariant individual characteristics.

Table 3 shows that both mood and subjective performance is higher (0.17-0.16 points higher) when working from home compared to when traveling to work. This result translates as a 3% shift in mood for the entire scale and a 3.7% increase relative to the average level (4.6). Similar interpretation yields for work performance.

When further excluding those who report having chosen freely to work from home (Step 2), the results remain true (0.12 and 0.06 points difference), but the effect size is lower. In particular, the effect for work performance is lower and loses significance when excluding those who chose WFH (Step 2).

The previous analysis does not allow us to further study whether the estimated effect of working from home is homogeneous or whether some people may be better or worse off when working from home. Given that several articles have previously found that individual characteristics play a role in the experience of working from home, we carry out an additional analysis to investigate this point.

To better understand if there are specific individual characteristics or personality traits that are important for mood and subjective performance in our sample, we run GLS panel regression only for those working from home. The regression models in Table 4 include several independent variables: demographics, transport and work-related characteristics, wellbeing and personality traits from the Big Five Inventory (see the description of variables in Section 2.3). Income and Education are two categorical variables with 6 and 4 levels respectively, and in the regressions, they are treated as continuous. The variable “commute>30min” and “Children < 18” are dummy variables. “Flexible work time” is also a dichotomous variable indicating whether the respondent had to be at the office at a specific time.

We see that today's positive experience with working from home is strongly associated with better mood (0,41) and higher performance (0,77). Scoring high in general wellbeing, emotional stability, being physically active and having a good work-life balance is also positively correlated with better mood and higher performance. Doing one's morning routines and some physical activity before starting to work is associated with reporting a better mood, while scoring high in consciousness is correlated with higher work performance.

Being female is associated with reporting slightly worse mood and performance (weakly significant), while older people tend to report slightly better mood and higher performance. Lower education, long commuting time and no previous experience with WFH is associated with better mood, while having a flexible job schedule seems to be negatively correlated with performance.

Interestingly, we do not find any effect for “Why WFH” (chosen, strongly suggested or imposed by the employer), hence it does not seem that choosing to work from home or having it imposed significantly impacted mood and performance in our sample. Moreover, income level and having children in the household also do not seem to have any impact on mood or work performance. These results are robust when controlled (using fixed effects) for time of response (May-Sep. 2020), place of residence and type of work.

As a consequence of the COVID 19 epidemic, teleworking became common in a number of jobs that had previously not considered this a relevant or possible option. A strength of the current study is that we included respondents from different jobs and organizations, in contrast with several previous studies that recruited participants from only one organization (e.g., Anderson et al., 2015; Bloom et al., 2015; Delanoeije and Verbruggen, 2020). Hence, our results extend evidence beyond single-organization studies, although the sample is not representative of the broader workforce. Our sample is composed mostly of high educated, high-earning women working in office-based or academic positions.

Given our voluntary recruitment through social media, the sample is not representative, and the results are therefore not generalisable to the whole population. Moreover, we surveyed only individuals who were still working and had jobs that could, at least partially, be performed from home. In the analysis, we further restricted the sample to those with work suitable for teleworking, excluding the 2% working in health services and the 1% in sectors such as farming, fishery, or transport.

In this study, we distinguish between two types of self-selection. First, individuals may choose to work from home because they have a positive experience with teleworking or a negative experience traveling to work due to Covid-19-related anxiety. Second, individuals who respond to the survey may do so because they have a positive experience with teleworking.

We argue that the first type of selection is mitigated by our panel structure and by exogenous changes in mandates to work from home that varied across place and time. Moreover, we do not observe an effect on mood or performance when teleworking was chosen vs. imposed by the employer (Table 4, Why WFH). However, the effect of working from home on work performance is reduced when excluding those who report choosing to work from home (Table 3, Step 2).

As with most other studies on teleworking, with some notable exemptions (e.g., Bloom et al., 2015), this study relies on self-reported data and does not employ an experimental design; therefore, the ability to draw causal conclusions is limited. The lockdown context reduced this challenge to some extent, as a large proportion of the sample were required to work from home, which allows us to treat the situation as a natural experiment. Nevertheless, not all participants transitioned to teleworking, and individual preferences may have influenced both the initial shift to working from home and the decision to return to the office during the re-opening phase. We address this by asking respondents whether teleworking was imposed or chosen, incorporating this information into the analysis, and conducting additional analyses excluding those who reported choosing to work from home.

In contrast, we are not able to remove self-selection into the survey itself due to the voluntary recruitment process. In addition, about 80% of the sample report that working from home functions quite well or very well for them. Hence, we cannot exclude that the study overrepresents individuals who are particularly satisfied with teleworking.

A strength of this study is the use of Experience-Sampling Method (Csikszentmihalyi and Larson, 2014) to elicit mood and work performance at each time point. The survey was sent to the participants mid-morning, and the instruction mail asked participants to answer a few hours after the start of their workday and before lunch, as well as that they should avoid/postpone responding if they had not started their workday. In doing so, we aim to ensure homogeneity of responses and avoid post-lunch dip. Moreover, removing self-choice in terms of response timing is also important when assessing transient states such as mood (Beute and de Kort, 2018). We observe an extremely high rate of compliance with instruction as about 90% respond mid-morning.

A limitation of the study is that some validated English-language scales required adaptation to Norwegian, including item reduction for certain constructs (e.g., mood) and the use of non-validated measures for others (e.g., performance). Although this approach is common in experience-sampling research, using a reduced item set limits formal claims regarding structural validity and sensitivity at the latent construct level. In addition, internal consistency for the personality scale was not always above the recommended threshold (α < 0.70 for three of five subscales). Personality was, however, used only as a control variable in parts of the analysis, limiting the impact of this issue on the primary results.

Our data were collected using a panel design, which is a strenght when it comes to the type of analysis methods available and it allows us to observe the same individuals over time. The study also shows quite low attrition rates (average response is 5.8 times out of 7), compared to what can be expected for online surveys during the workday.

In contrast with other studies, we do not ask participants to relate their experience of WFH to their work experience before Covid-19 to avoid recollection bias and emotion anchoring to before the pandemic situation. In contrast, we assess the impact of WFH, versus traveling to work, on mood and performance in a period which many have referred to as the “new normal” providing more realistic and relevant insights for future work organization. A limitation of such approach is that traveling during the Covid-19 pandemic may be associated with negative emotions such as stress or anxiety from the possibility of contagion. We control for such possibility by including a measure of Covid-19 anxiety in our subgroup analysis.

Finally, a limitation of our survey is that we don't know how often people worked from home overall. The follow-up surveys only asked about the given day, and someone who worked from home every time they answered the survey may have worked at the office other days of the week, or vice versa. Previous research has found that job satisfaction and the impact of professional isolation on job performance may depend on the extent of teleworking (Golden and Veiga, 2005; Golden et al., 2008), and we are not able to control for that.

Our main research question focuses on the impact of working from home, contra traveling to the office, a measure that was imposed as a consequence of the Covid-19 pandemic in 2020. Our survey collected information on about 2000 Norwegian workers, and it was conducted in the aftermath of the first Covid-19 lockdown. We observe the sample between May and September 2020 during a gradual re-opening of society where the amount of people working from home changed over time and place.

Between March to May 2020 (during the first lock down), about 88% of our sample report having exclusively worked from home and that it worked quite or very well for them. In our sample, about 9% exclusively traveled to work, while 35% reports always working from home. The majority (56%) reports both traveling to work and working from home during the observation period.

Our results show that people who worked from home between May and September 2020 reported a better mood and higher work performance compared to when traveling to work. Positive experience with teleworking on the current day is a strong predictor for both mood and performance, as well as general wellbeing, emotional stability, being (generally) physically active and having a good work-life balance. Doing one's morning routines or being physically active before work seems to be associated with better mood, while those who score high in consciousness tend to report higher work performance. One conclusion is that any type of physical activity (active travel or training) before work may play a role for people's mood and performance during the day.

This study is novel and contributes to the scientific literature by reporting effects of working from home through exploitation of exogenous geographical and time variation of work location due to Covid-19 travel restrictions in a panel survey.

Our results provide important inputs for employers, policy design and for future labor organization. Teleworking is particularly relevant for people who have a long commute. Working from home could be encouraged when the conditions allow for it i.e., suitable work activities and good enough conditions that allow for a positive experience of working from home. The findings underline the importance of maintaining healthy routines such as physical activity and prioritizing a good work-life balance, as they have positive impacts on mental health and work performance. From a broader policy perspective, teleworking also interacts with transport and sustainability goals. Reduced commuting demand can ease congestion and lower emissions, particularly in urban regions, while still supporting worker wellbeing. All in all, even modest improvements in daily mood and subjective performance associated with working from home may accumulate over time, suggesting that remote work can contribute positively to worker wellbeing when implemented under supportive conditions.